| Title: | Methodology for analyzing ranges of uncertain model parameters and their impact on total maximum daily load processes |
| Authors: | Benaman, J. and C.A. Shoemaker |
| Year: | 2004 |
| Journal: | Journal of Environmental Engineering |
| Volume (Issue): | 130(6) |
| Pages: | 648-656 |
| Article ID: | |
| DOI: | 10.1061/(ASCE)0733-9372(2004)130:6(648) |
| URL (non-DOI journals): | |
| Model: | SWAT |
| Broad Application Category: | hydrologic and pollutant |
| Primary Application Category: | calibration, sensitivity, and/or uncertainty analysis |
| Secondary Application Category: | Total Maximum Daily Load (TMDL) applications |
| Watershed Description: | 1,178 km^2 Cannonsville River, located in southeast New York, U.S. |
| Calibration Summary: | |
| Validation Summary: | |
| General Comments: | A method for assessing uncertainty in parameter ranges is described and applied, that uses both Monte Carlo simulations and interval-spaced sensitivity analysis. One goal of the approach is to ascertain portions of parameter ranges that can result in unrealistic model estimates. An initial set of 500 Monte Carlo simulations were performed with 36 different input parameters; 22 of the parameters were analyzed in a subsequent interval-spaced sensitivity analysis (22 parameters were those that directly affect erosion or sediment transport). Resulting spider plots indicate that some parameters have very large ranges that significantly affect the model output (e.g., SPCON has range exceeding 5,000% of base value). The 500 Monte Carlo simulations were executed again for the 36 parameters, but reduced parameter ranges were used for 3 key variables. The authors further discuss the implications of their results for TMDL applications. |
| Abstract: | A methodology for analyzing uncertain parameter ranges prior to model calibration or uncertainty analysis is presented. The method considers parameters that exist in complex models and are typically difficult to set using site-specific data (i.e., parameters that have suggested ranges, national average ranges, or ranges set with land characteristic data). The method applies Monte Carlo runs and an interval-spaced sensitivity analysis to determine the parts of parameter ranges that will most likely cause unrealistic model results. An application of the method is presented using the Soil and Water Assessment Tool model as applied to the Cannonsville Reservoir system watershed for hydrology and sediment simulations. Results indicate that after parameter range reduction, the model output range was reduced by an order of magnitude, thereby reducing the uncertainty of the model and aiding the calibration effort. Sediment transport is difficult to monitor and model in its many stages of transport so significant uncertainty in the sediment erosion and transport parameters for this model still exist. This uncertainty will impact the application of the model for Total Maximum Daily Load development and management decisions. |
| Language: | English |
| Keywords: | Modeling; Water quality; Water pollution |